PlacementOS

AI-orchestrated campus placement automation on UiPath Maestro BPMN. Turns 2 weeks of manual resume screening, scheduling, and offer dispatch into a 48-hour automated end-to-end process.

Comment

Inspiration

Every year, 1.5 million engineering students in India go through campus placements. The Training & Placement Officer running each drive does it manually — 500+ resumes screened by hand, interview slots coordinated over WhatsApp, offer letters typed in Word. A single drive takes 2 weeks. PlacementOS automates it to 48 hours.

What it does

PlacementOS orchestrates the entire campus placement drive lifecycle as a 10-step BPMN 2.0 process on UiPath Maestro:

  1. Fetch Eligible Students — RPA bot pulls students from the college ERP filtered by CGPA
  2. AI Resume Screening — LangChain + Groq (Llama 3) scores each candidate 0–100 against the Job Description, with reasoning, strengths, and gaps
  3. TPO Shortlist Review — Training & Placement Officer reviews the AI shortlist via UiPath Action Center (human-in-the-loop gate)
  4. HR Interview Slot Confirmation — HR confirms available slots via Action Center
  5. Calendar Invites & Email Dispatch — RPA bot sends calendar invites and confirmation emails
  6. Interview — Conducted externally
  7. HR Submits Results — HR logs interview outcomes via Action Center
  8. Offer Letter Generation — API workflow generates offer/rejection PDFs
  9. Offer Dispatch — RPA bot emails offer letters to selected candidates
  10. ERP Record Update — Placement outcomes written back to the college ERP

The process includes 2 exclusive gateways (no candidates found → drive cancelled; shortlist rejected → drive cancelled), a 24-hour SLA timer on the TPO review task, and 3 terminal end events.

How we built it

UiPath Maestro BPMN is the orchestration spine — the entire 10-step process runs as a live BPMN 2.0 process on UiPath Automation Cloud (Track 2).

External AI Agent — A FastAPI microservice built with LangChain + Groq (Llama 3, free tier) handles resume screening. The Maestro process calls this agent via HTTP; Groq scores candidates asynchronously in batches.

Mock ERP API — A FastAPI service on Render (Singapore) simulates the college ERP with 20 realistic BITS Pilani students across CSE, IT, ECE, EEE, and Mechanical branches, with CGPAs ranging from 6.2 to 9.3.

UiPath Orchestrator Assets — 5 assets connect the live services: ERP_BASE_URL, SCREENER_URL, SMTP_HOST, SMTP_PORT, SMTP_USER.

RPA Bots (UiPath Studio) — ERPFetchBot, CalendarEmailBot, and PlacementRecordBot handle ERP calls, email dispatch, and record updates.

Action Center — Human-in-the-loop tasks for TPO shortlist approval and HR slot confirmation are routed through UiPath Action Center with structured form schemas.

Claude Code (UiPath for Coding Agents) — Every component was built using Claude Code as the coding agent. Claude Code generated the FastAPI backends, BPMN 2.0 XML (500+ lines), RPA bot XAML scaffolding, Action Center form schemas, and deployment config from natural language prompts. See docs/claude-code-demo.md in the repo for exact prompts used.

Challenges we ran into

  • BPMN timer boundary events — Configuring the 24-hour SLA escalation timer on the TPO task required careful non-interrupting boundary event configuration in UiPath Studio Web
  • Groq rate limits — Implemented exponential backoff in the resume screener to handle Groq API rate limits during batch processing of 20+ resumes
  • Action Center form schemas — Designing structured JSON forms for the TPO and HR tasks so that human inputs flow cleanly back into the BPMN process variables took several iterations
  • Render cold starts — Both APIs are on Render free tier; added retry logic in the BPMN service tasks to handle cold start latency

Accomplishments that we're proud of

  • A fully working 10-step BPMN process with 2 successful end-to-end runs on UiPath Automation Cloud
  • Live AI resume screener scoring real candidates with detailed reasoning (top candidate: CGPA 9.1, Microsoft IDC internship, AAAI 2024 paper → Score 95, STRONG_YES)
  • An entire enterprise-grade system built using Claude Code as the coding agent — demonstrating UiPath for Coding Agents in a real workflow

What we learned

  • UiPath Maestro BPMN is genuinely powerful for process orchestration — the ability to mix RPA bots, external AI agents, API calls, and human tasks in a single BPMN process is enterprise-grade
  • Groq's free tier with Llama 3 is surprisingly capable for structured resume scoring tasks when prompted well with LangChain
  • Claude Code dramatically accelerates backend development — what would have taken days of boilerplate took hours of natural language iteration

What's next for PlacementOS

  • Replace mock ERP with real ERP integration (SAP, Oracle) via UiPath connectors
  • Add Google Calendar API integration for real slot booking
  • Expand to cover pre-placement activities: company registration, student registration, eligibility verification
  • Deploy to a real college placement office for a pilot drive

Built With

  • claude-code
  • fastapi
  • groq
  • langchain
  • llama-3
  • python
  • render
  • uipath-action-center
  • uipath-maestro-bpmn
  • uipath-orchestrator
  • uipath-rpa
  • uipath-studio
Share this project:

Updates

Submission history